Successfully reported this slideshow.
We use your LinkedIn profile and activity data to personalize ads and to show you more relevant ads. You can change your ad preferences anytime.

Tying, Entry, and Competition in Investment Banking


Published on

  • Be the first to comment

  • Be the first to like this

Tying, Entry, and Competition in Investment Banking

  1. 1. Tying, Entry, and Competition in Investment Banking Christian Laux Uwe Walz J.W. Goethe-University Frankfurt J.W. Goethe-University Frankfurt∗ March 2004 Abstract We analyze the role that tying credit and fee business has for the entry and compe- tition in investment banking. Specialized investment banks that offer only investment banking services earn a rent that induces them not to renege on their promise to offer high quality service. Entering this lucrative market is impossible for firms that (ini- tially) have higher costs of providing services because the higher costs result in higher incentives to shirk. Tying provides additional incentives as the value of risky debt is adversely affected if low quality advice is provided. Tying allows to enter the invest- ment banking business but at the same time leads to stiffer price competition, reducing the profitability of this business. JEL Classification: G21, G24, D49 Keywords: tying, investment banking, universal banking ∗ Address of authors: Christian Laux, Department of Finance, Mertonstr. 17, 60325 Frankfurt, Ger- many, e-mail:; Uwe Walz, Department of Economics, Schumannstr. 60, 60325 Frankfurt, Germany, e-mail: 1
  2. 2. 1 Introduction In the last decade an ongoing process of integration between investment and commercial banking has been observed in the banking sector of different countries. Most notably, this has been the case in the US where this process was facilitated by the gradual relaxation of the legal barriers between the two sectors of the banking industry. Until the mid 1990s the Glass-Steagall act was the main barrier for combining investment and commercial banking in the US. Therefore, firms in the world’s main capital market were not able to purchase the services of commercial and investment banking from a single bank. On the same grounds, commercial banks were not able to enter the often considered highly profitable investment banking sector (and vice versa). While the sharp division of the Glass-Steagall act between the two sectors de facto has been hollowed out already earlier, it took until 2000, until the US Congress abolished the act entirely. There is ample empirical evidence that since the 1990s, the degree of tying of fee-related business of investment banks (underwriting equity and bond issues, M&A etc.) and credit business of commercial banks has substantially increased (Drucker and Puri (2003)). Em- pirical studies and stories in the popular press (see, e.g., Cairns et al. (2002)) suggest that there are two directions. On the one hand, universal banks tend to use their balance sheet (i.e., their capability to provide loans to firms) to acquire market share in the investment banking sector: universal banks demand fee business in return for credit. On the other hand, some clients seem to demand credit in return for fee business. This leads us to the two main questions of our analysis: 1. Why does tying occur? 2. Who benefits from tying and what are the consequences of it? The starting point of our analysis is an incentive problem in investment banking. To provide investment banks with incentives to exert costly effort to provide high quality service, 2
  3. 3. they have to earn a rent. The rent stems from high fees that are paid for investment banking services. If an investment bank shirks on the quality of its advice, it stands to lose its reputation, future business, and rent that goes along with it. The existence of this incentive problem and the resulting rent make the investment banking business lucrative and worthwhile for commercial (universal) banks to enter even if they (initially) have higher costs of providing high quality service. However, the problem is that even if it is worthwhile for entrants to provide service at the same price as incumbents, it is not possible to enter the investment banking market. The reason is that the higher costs of providing service increase incentives to shirk on the quality. Firms foresee that an entrant would shirk on the quality of advice and therefore not obtain advice from an entrant. Providing an investment banking client with credit tightens incentives to provide high quality advice: as the quality of advice affects firm value it also affects the value of risky debt; shirking on the quality of advice reduces the value of the outstanding debt. A universal bank can therefore use tied deals to tighten incentives and to enter the investment banking market. This is the bright side of tied deals. However, there is also a dark side for providers of investment banking services. With tied deals it is possible to reduce the advisory fee as debt substitutes for rent in providing incentives. Therefore, competition reduces the rent to be earned in investment banking in tied deals. A lower bound is set to the rent only by constraints on how much debt can be provided. We derive the following empirical implications: (i) Tying is particularly important for universal banks that have a cost disadvantage over specialized investment banks. (ii) Firms that obtain tied deals have higher debt levels as a high debt capacity is required for debt to provide sufficient incentives. (iii) Tying leads to more aggressive pricing, reducing the profitability of the fee business. (iv) Tying can be used to provide advice when its marginal profitability is lower than the rent that induces specialized investment banks to provide high quality advice. (v) Universal banks are more willing to provide risky debt (or credit 3
  4. 4. lines) if this allows them to sell fee business. (vi) Markets for investment banking services are segmented with specialized investment banks providing services for which they have a comparative (cost) advantage to firms with low debt capacity, and universal banks offer tied deals for ”standard” service provided to firms with high debt capacity. Our paper is closely related to an extensive literature that analyzes the implications of commercial banks entering the investment banking business. (See Rajan (1996) for a survey.) A dominant theme in this literature are information externalities: banks monitor firms to which they provide credit, which gives them a comparative cost advantage in also providing information-sensitive fee business. But information spillover can also be a source for potential conflicts of interest, that are, for example, discussed by Benston (1990), Kroszner and Rajan (1994), Puri (1994, 1999), and Kanatas and Qi (1998). In our paper there is an incentive spillover: debt provides incentives to offer high quality advice; this can provide a competitive advantage even if the costs of producing the service are higher as it aligns incentives. Kanatas and Qi (2003) show that a universal bank’s incentives to exert underwriting effort might be low because, if underwriting fails, the bank has a comparative advantage in providing credit to the firm as it already invested in acquiring information about the client. In contrast, a specialized investment bank is not able to provide credit if underwriting fails and has higher incentives to avoid failure. Kanatas and Qi do not address tying. Instead, it focuses on credit and underwriting as substitutes: credit is used at then prevailing conditions if underwriting fails, which reduces a universal bank’s incentives. In contrast, we discuss the situation where credit and investment banking service are both provided simultaneously. In this case tying (i.e., providing credit or a credit line at fixed terms by the underwriter at the time or prior to underwriting) increases the universal bank’s incentives. The recent empirical analysis of Drucker and Puri (2003) can be regarded as an empirical ”complement” to our paper. By using a data set for the 1996-2001 period they analyze the empirical relevance of tying as well as the manner in which tying occurs. Furthermore, they 4
  5. 5. analyze the consequences of tying on the parties involved as well as on financing costs. They clearly show the empirical relevance of tying. The relation between their empirical results and our analysis is discussed later in this paper. Chemmanur and Fulghieri (1994) were the first to discuss the implications on reputation concerns in investment banking. We choose a simpler model of reputation in investment banking as a starting point to analyze incentives stemming from credit provision as a com- plement to reputation. There are also some obvious links between our approach and the literature on tying in the industrial organization literature. A number of papers in that area provide arguments that a policy of a incumbent supplier of two complementary products may deter entry by making the prospects of entry less attractive and certain (see Choi and Stefanidis (2001)).1 In our paper, tying, just has the opposite effect: it serve as an entry-inducing device. The main mechanism of our paper, namely that lending acts a bonding mechanism which prevents shirking is reminiscent to analyses in the labour market literature (see e.g. Lazear (1979) and Allen et al. (1993)). The paper is organized as follows. In the next section we outline our main set-up. There we describe the incentive problem on which our analysis is based. In section three we analyze the competition among investment banks alone and look into the possibility of entry of a commercial bank in the absence of tying. The fourth section is devoted to an analysis of tying and entry of commercial banks in the absence of capacity constraints. The fifth section introduces upper limits on the provision of risky debt by commercial banks and looks into the consequences of this on market equilibrium. The sixth section discusses the empirical implications of our model. In section seven we discuss the robustness and extensions. 1 Other papers stress the potential of tying as an entry-deterring device, as well (see e.g. Whinston (1990) and Carlton/Waldman (2002)). 5
  6. 6. 2 Model There is an infinite number of periods and three types of players. First, in each period there is a large number of owners (firms) with access to new projects. To realize their projects, firms need financing; in addition, they may seek investment banking advice. Second, investment banks are specialized in advising clients on IPOs, M&As, financing decisions, etc. (fee-business). Third, commercial banks are specialized in lending (credit business). But commercial banks might also consider entering the fee-business (universal banks). All parties are risk neutral and the risk-free rate of return is r. Owners can seek advice from a specialized investment bank or the investment bank branch of a universal bank to increase expected profits. Projects and advice: We consider the simplest setting possible. Each project requires an investment I and realizes either a high payoff π (success) or a low payoff π (failure), with π > I > π. The success probability θ depends on the quality of advice provided by the investment bank. There are only two quality levels, high and low, θ ∈ {θh , θl }, with θh > θl . θl can be realized without an investment bank. Providing high quality advice is costly for investment banks while the costs of choosing low quality are normalized to zero. High quality is not verifiable by a court. Therefore, incentives have to be provided to the ”investment bank” to choose high quality advice. This stylized incentive problem is at the heart of our analysis. For convenience it is assumed that investment, advice, and profit realization all occur at the same time. Therefore, there is no discounting within periods. Projects have a positive net present value even without advice, i.e., given θl . Costs of high quality advice: Specialized investment banks’ cost of high quality advice is denoted by c (per unit of advice). Commercial banks that want to enter investment banking can produce the same quality of advice. But their costs of giving high quality 6
  7. 7. advice are initially cU and strictly higher than the costs of specialized investment banks, ¯ c. But eventually the costs fall to cU ((1 + r)c > cU ≥ c). The differences in the costs of producing high quality advice might be due to learning, initial setup cost, specialization, etc. It is assumed that high quality advice is advantageous for firms even if produced at cU , i.e., ¯ θh π + (1 − θh )π − cU > θl π + (1 − θl )π. ¯ Rearranging yields ∆(π − π) > cU , ¯ (1) with ∆ ≡ (θh − θl ). Financing: The firm has no funds of its own. It can finance the project with debt and equity. We do not model the cost of providing credit but assume that these costs are (much) higher for investment banks than for commercial banks. It is therefore very costly for investment banks to provide loans. Both types of banks are assumed to have prohibitive costs of investing equity on their own. However, high-quality advice from an investment bank enables the firm to raise external equity. Debt and external equity are competitively priced and the expected repayment equals the amount raised. Hence, assuming a firm raise B through bank borrowing and E through issuing equity, the debt repayment obligation D ≤ π is B = θi D + (1 − θi ) min{π, D} and the share α ≤ 1 in the firm’s equity is E = α[θi (π − D) + (1 − θi ) max{0, π − D}], where θi is the equilibrium choice of advice quality that debt holders and equity holders correctly anticipate. To finance the project, the firm has to raise B + E ≥ I; excess capital, B + E − I > 0, is paid out to initial owners. 7
  8. 8. 3 No tying: independent financing and investment ad- vice The base case of the model is set up such that there is no difference between debt and equity for the firm if there is no tying. The firm obtains either debt or equity independent of the quality of advice. That is, there is no inherent advantage of one source of financing over the other. Of interest are the incentives for providing high quality advice. Incentives and ”reputation” in investment banking: We first analyze the case where there are only specialized investment banks. The chosen quality is not contractible and explicit monetary incentive schemes are not possible. Investment banks’ incentives to provide high quality advice stem from their reputation. If they do not provide high quality advice, they put themselves at the danger of not obtaining future business. Hence, it is essential that investment banks earn a rent that is at stake. To model the equilibrium we assume Bertrand competition between N investment banks for mandates of M firms. The sequence of events in each period is as follows: 1. All investment banks simultaneously quote a price p at which they are willing to offer advice. 2. Firms choose whether and from which investment bank to obtain advice; mandates are equally distributed if they are indifferent. 3. Investment banks choose whether to provide high or low quality advice for the indi- vidual projects for which they provide service. 4. ”The market” observes if an investment bank shirks and provides low quality advice. The assumptions are chosen to carve out the main arguments in the simplest setting possible. They are stronger than necessary. For example, there might be some ambiguity 8
  9. 9. about the quality of advice and low quality advice might be detected with a probability lower than one. Moreover, we can also allow for monetary incentives as long as they do not eliminate the investment banks’ rent. We find: Lemma 1 The following constitutes a subgame perfect Nash equilibrium: • All investment banks offer advice at price p∗ = (1 + r)c and choose high quality. • If (θh − θl )(π − π) ≥ (1 + r)c, firms seek advice at price p∗ from an investment bank that has a history of consistently providing high quality advice. If (θh − θl )(π − π) < (1 + r)c, firms carry out the project without advice. To derive the equilibrium, we show that the strategies are indeed optimal for investment banks and firms. When deciding on the optimal level of advice investment banks consider their expected payoffs from either providing low or high quality. Since providing low quality on one project results in zero demand for advice in the future, it always pays to shirk on all mandates (or not at all).2 Let m be the number of customers that each investment bank obtains in equilibrium. (Ignoring integer problems, m = N/M .) With high quality advice, an investment bank’s total profit is m(p − c) + m(p − c)/r, where the latter term denotes the present value of the investment bank’s rent from future business. With low quality the investment bank’s payoff is mp. Hence, for m(p − c) + m(p − c)/r ≥ mp, it is optimal for the investment bank to provide high quality advice. Rearranging yields p−c c≤ , (2) r 2 If instead we assumed that shirking is observed with a positive probability that is increasing in the number of projects on which the investment bank shirked, the investment bank might consider shirking on less than all projects for which it provides advice. 9
  10. 10. Firms only seek advice if they expect investment banks to provide high quality (i.e., (2) holds) and if the expected increase in profits exceeds the price for advice (i.e., ∆(π − π) ≥ p). Moreover, they will seek advice from the investment bank that offers high quality advice at the lowest price. p∗ = (1 + r)c is the lowest price for which investment banks have an incentive to choose high quality. Any deviation from p∗ by individual investment banks results in zero demand. It is immediately clear that it is not optimal for a firm to choose an investment bank that offers advice at a price exceeding p∗ . But it is also not optimal to choose an investment bank that offers advice at a price below p∗ because this bank will not choose high quality. The only reason for undercutting in the current period p∗ (which has no direct implications on the incentive of the bank) is to increase demand to m > m. But higher demand is only sustained ¯ in future periods if the price remains low, jeopardizing incentives because the future rent is too low. However, obtaining a higher demand only once (and returning to the equilibrium price p∗ after one period) also jeopardizes incentives because the saving from not offering −c ∗ high quality in this period exceeds mc if the demand exceeds m, i.e. mc > m( p r ) = mc. ¯ We assume that ∆(π − π) ≥ (1 + r)c. Therefore, firms benefit from obtaining advice at price p∗ and investment banks have no incentive to shirk on the quality. Entry of commercial banks A commercial bank enters the market by quoting a price at which it is willing to offer advice. If customers are indifferent between the entering commercial bank and the investment banks, the total demand is split evenly between them. In this section we assume that the entering commercial bank has a cost disadvantage only in the first period. That is, cU > c at the time of entry and cU = c in all future periods. By ¯ showing that even in this extreme case in which the potential entrant has a cost disadvantage just in one period, allows us to carve out our main argument as clearly as possible. In order to compete with investment banks, the commercial bank must not charge a higher price than p∗ . If entry is successful, the future rent from doing business is the same 10
  11. 11. for the investment bank branch of the universal bank as for specialized investment banks, as cU = c . Entry is profitable for the entrant if p − cU + (p∗ − c)/r = p − cU + c > 0. ¯ ¯ (3) Note that we implicitly assume that the entrant has to take on his share of the market in the first period, resulting in total costs me cU in the first period and a rent from future operations me (p∗ − c)/r, where me is the number of customers for each investment bank after entry. In particular, it is not possible to service only one customer in the first period when cost of production are high. This captures the idea that learning, resulting in lower costs of providing high quality advice, requires sufficiently many customers. We assume that (3) is satisfied. That is, a commercial bank is willing to offer advice at the same conditions as investment banks and to invest the higher costs of providing high quality advice in the first period. However, the commercial bank is unable to enter the market. Proposition 1 Assuming that only the rent from future business provides incentives to choose high quality, entrants will not receive any business and cannot enter the market char- acterized in Lemma 1. Proof: A commercial bank cannot enter the market due to its higher costs of producing high quality advice at the time of entry, which provides the entrant with higher incentives to shirk. Comparing the cost savings in the case of shirking with future gains when providing high quality advice yields: p∗ − c cU > ¯ = c. r Note that again it is not possible to service only one customer in the first period when cost of production are high, to reduce incentives to shirk, and then service me customers at cost c in the next period. 11
  12. 12. 4 The role of combining credit and investment banking business 4.1 A single commercial bank For a commercial bank it is possible to enter the investment banking business by tying credit provision and investment banking advice. Proposition 2 Assuming that the demand for advice is positive at p∗ = (1 + r)c, a com- mercial bank can enter the market by simultaneously offering to provide investment advice at price p∗ and credit with a repayment obligation D, where D satisfies D > π and cU ≤ (θh − θl )(D − π) + c. ¯ Proof: Providing credit changes the entrant’s expected payoff conditional on the quality of his advice. The expected payoff is p + θl D + (1 − θl )π if he shirks and p − cU + θh D + ¯ p∗ −c (1 − θh )π + r if he provides high quality advice. Comparing the two expressions for p∗ = (1 + r)c reveals that the entrant has an incentive to provide high quality advice if the inequality in the Proposition holds, thereby enabling the commercial bank to enter. By tying credit provision and advice, the universal bank can increase incentives not to shirk on the quality of its advice. For D > π, the expected debt repayment decreases by ∆(D − π) if the universal bank provides low quality advice. This provides additional incentives to offer high quality advice. Firms are indifferent between (a) buying advice from a specialized investment bank and obtaining financing in the market and (b) buying advice from a universal bank that also provides it with credit with a repayment obligation D as characterized in the Proposition. 12
  13. 13. The analysis generalizes to future periods: tying is necessary whenever the universal bank has higher costs of producing high quality. Therefore, if cU > c, tying remains to be necessary for the universal bank even after entry. Because of the higher costs, the universal bank has higher incentives to shirk on quality and therefore cannot offer investment banking advice on a stand alone basis. Despite the higher cost of producing quality advice, it is worthwhile for the commercial bank to enter the market, because of the rent that can be earned, if p∗ − cU + (p∗ − cU )/r > 0. ¯ Note that we implicitly assume hat the entry of a commercial bank in the fee business is not foreseen by specialized investment banks. If investment banks know that a commercial bank will enter the market next period, they will shirk on the quality of advice because the entry reduces the number of mandates that each can sell in the future. This in turn reduces the future rent and therefore the investment bank’s advantage of shirking exceeds the future rent.3 The Proposition establishes that entry in investment banking is possible through tying. It does not characterize the universal bank’s optimal strategy. In particular, it was assumed that the universal bank offers investment banking advice at the same price p∗ as specialized investment banks. If specialized investment banks deviate from this price, they will obtain zero demand. This is not true for investment bank branches of commercial banks. Of course, the commercial bank cannot sell services at a higher price than p∗ . But it can offer advice at a price below p∗ in tied deals to obtain a higher share of the investment banking business. We can thus state: 3 Assuming a constant probability of market entry, β, would not change our qualitative results but lead to an increase in the equilibrium price. In this case the incentive constraint is mc ≤ (1 − β)m(p+ − c)/r + βmi (p+ − c)/r, where mi < m is the future demand for each investment bank after entry. If the probability of entry, β, increases, the price p+ must increase as well. 13
  14. 14. Lemma 2 Without any constraint on the amount of debt that the commercial bank can provide and with only one commercial bank, the commercial bank will use tying and a price marginally lower than p∗ to monopolize the market. The Lemma is not meant to be taken literally. Instead, it shows that it is indeed possi- ble to use tying to obtain a competitive advantage. But in the present context tying does not result in higher prices. Tying is akin to introducing an additional (superior) incentive mechanism. Therefore, two questions arise. First, what happens if more than one commer- cial bank enter the market and there are multiple competitors with access to the incentive mechanism? Second, when can universal banks and specialized investment banks (that do not provide credit) coexist? We approach these two issue in the following. 4.2 Multiple competing commercial banks and no ”capacity con- straint” If there is no constraint on the amount of debt that each commercial bank can provide, we can, without loss of generality, restrict attention to two commercial banks. Lemma 3 Assume that there is no constraint on the amount of debt that each commercial bank can use in a tied deal. 1. If there is one commercial bank that entered the market first, this bank offers advice at price being marginally below p = min{¯U , p∗ } in each period after entry, and the second c commercial bank never enters. 2. If both commercial banks want to enter the market simultaneously and cU ≤ p∗ , both ¯ banks offer advice at price p = cU at the time of entry and p = cU thereafter. ¯ 3. If both commercial banks want to enter the market simultaneously and cU > p∗ , both ¯ 14
  15. 15. banks randomize. In the mixed strategy equilibrium banks enter the market with prob- ability x = 1 − ((¯u − p∗ )r)/(p∗ − cu ). c In the following we derive the lemma. (1) If one commercial bank has entered the market it can monopolize the market with a limit-pricing strategy. Investment banks cannot offer advice at a price below p∗ and the second commercial bank has no incentive to enter the market at a price below cU . Therefore, a price marginally below min{¯U , p∗ } keeps both out ¯ c of the market. (2) When both commercial banks enter the fee business, Bertrand competition leads to (marginal) cost pricing in all periods. (3) Given cU > p∗ , commercial banks have to charge a price below cU for advice at the time ¯ ¯ of entry. But if both enter, their future rent is zero because of Bertrand competition, making it not profitable for both to choose a price below cU to enter. Hence, they will randomize ¯ between ”entering the market and offering advice at price p∗ ” and ”not entering”. If only one offers advice, the continuation game is as characterized in part (1) of the lemma, market entry is profitable for this bank. If both enter, they will compete prices down and the costs of entry into the investment banking business (offering advice at a price below the cost of producing the advice) are not recovered through future rents. Due to randomization it may happen that neither of the two banks enter, implying that the continuation game is just a replication of the present period’s situation. The detailed derivation of the mixed strategies used is delegated to the appendix. Of course, the setting is rather stylized. But we believe that the main insights generalize: Tying credit and advice allows commercial banks to enter the market. This is the bright side of tied deals. However, competition between commercial banks leads commercial banks to use tied deals in the future as well and to lower prices for advice. This price reduction that is possible through tied deals reduces the profitability of the investment banking business. Indeed, it can result in situations where entry in the market results in a loss. This is the 15
  16. 16. dark side of tying. We will discuss two potential ”remedies” to this dark side of tying in the next chapter. 5 On the coexistence of tied and non-tied deals Without any constraint on the maximum amount of debt that a commercial bank can use in tied deals, it can always ensure that its incentive constraint is fulfilled: by choosing D = π the commercial bank internalizes the effect of high quality advice. Since we assume (1 + r)c > cU , the commercial bank is able to capture the entire market by choosing a price below p∗ = (1 + r)c. We will depart from this extreme setting and look into the consequences of possible constraints on the amount of debt that can be used (either because individual projects have a debt capacity or because commercial banks have a constraint on how much risky debt they can lend). 5.1 Constraint on the maximum debt level for individual projects A high level of risky debt provides incentives to the commercial bank but will go along with increased costs of financial distress and distorted incentives for the firm. Therefore, it is costly to finance projects with very high levels of debt. We model these costs as a risk shifting problem that constrains the amount of debt that can be used in tied deals. We now assume that there is a second type of project that the firm can choose. The alternative project realizes a higher payoff in the case of a success, π 2 (π 2 > π). But it has a lower expected payoff, due to a lower probability of success. The success probability of the alternative project is given by qθi (i = h, l), with q < 1. Indeed, we assume that the alternative project has a negative net present value (even with high quality advice) and that it is not worthwhile to finance this project (even for an entrant in the investment banking business). Therefore, the debt level must be low enough not to provide incentives for the firm to 16
  17. 17. choose project 2, i.e., D must satisfy θh (π − D) ≥ qθh (π 2 − D) and the maximum debt repayment where the firm still chooses project 1 is given by ˆ D = (π − qπ 2 )/(1 − q). (4) We assume in the following that risk shifting does not impede debt financing altogether, i.e. ˆ D > π always holds. ˆ D constitutes an upper limit on the debt level that can be used in tied deals. With such an upper limit we will observe potential market segmentation. Firms for which the debt capacity is sufficiently high will be natural clients of universal banks providing tied ˆ deals. Firms which have a low D are typically advised by investment banks (and seek equity financing). We will investigate this market segmentation in more detail now. 5.1.1 Market segments Projects may now differ in two dimensions, their success probability’s sensitivity to high ˆ quality advice, ∆ ≡ (θh − θl ) ∈ (0, 1), and the maximum debt capacity D ∈ (π, π) (stemming from differences in the risk shifting problem). The number of projects with a given set ˆ ˆ of parameters (∆, D) remains constant across periods. The characteristics (∆, D) of each project are observed by the market. All types of projects have a positive net present value even without advice, i.e., with θl . (We assume that there are sufficiently many firms for which advice is efficient even if produced at high costs, i.e., for which (1) holds.) If firms expect high quality advice to be provided, they demand advice at price p if ∆(π − π) ≥ p. Without tying the market price is p∗ = (1 + r)c and specialized investment banks can service firms for which ∆ is sufficiently high, i.e., ∆(π − π) ≥ (1 + r)c (5) 17
  18. 18. Therefore, some firms, with a low expected increase in profitability due to advice, will not seek advice because of the high price that has to be paid. ˆ Universal banks can offer advice at price p to firms for which (cU , θh , D) satisfies ˆ cU ≤ ∆(D − π) + (p − cU )/r. (6) But firms will only seek advice from commercial banks at a price p ≤ p∗ = (1 + r)c. Substituting p∗ and rearranging yields ˆ (1 + r)(cU − c)/r ≤ ∆(D − π). (7) (7) is a necessary condition for universal banks to be able to compete with specialized ˆ investment banks for customers with characteristics (cU , ∆, D). The left hand side of the condition stems from the higher incentive to shirk at price p∗ due to its higher costs of producing high quality advice. The right hand side resembles the higher incentive to provide ˆ high quality advice with a tied deal for which D = D. Clearly, if cU = c, universal banks can compete with investment banks without using tied deals, i.e., the investment banking branch can survive on its own. But tied deals allow commercial banks to undercut the price of specialized investment banks. Whenever a universal bank has a cost disadvantage over a specialized investment bank in producing high quality advice, i.e., cU > c, the universal bank has to offer advice through tied deals. A universal bank’s ability to compete with specialized investment banks depends ˆ ˆ on the parameters (cU , ∆, D): it decreases in cU , increases in ∆, and increases in D. We can distinguish four market segments. ˆ Segment 1: All combinations of (cU , ∆, D) for which (5) and (7) are satisfied. This market segment will be taken over by the universal bank. It involves fee business from firms with sufficiently high debt capacity for which the universal bank’s cost disadvantage is not too high and that has a large impact on profitability. ˆ Segment 2: All (cU , ∆, D) for which (7) is satisfied but not (5). This market segment does not seek advice from specialized investment banks because of the high price required 18
  19. 19. to induce investment banks to provide high quality. A universal bank can use tied deals to reduce the price while retaining incentives to provide high quality. Therefore, if a firm’s debt capacity is sufficiently high, universal banks can provide these firms with advice that has low impact on profitability (low ∆) provided that the cost of providing this advice is not too high. This will be ”standard advice with low marginal profitability” but that potentially makes up a large fraction of the market potential. As it is likely that individual firms have several occasions where they would use advice but where advice profitability is low, servicing this market segment will lead to repeated business at lower price in tied deals. ˆ Segment 3: All (cU , ∆, D) for which is satisfied (5) but not (7). In this segment universal banks cannot compete with specialized investment banks. This segment involves deals for which the cost difference cU − c (i.e., the comparative advantage of investment banks) is rather pronounced to firms with low debt capacity. Examples include specialized advice to start-up firms and IPOs. ˆ Segment 4: All (cU , ∆, D) for which (5) and (7) are both not satisfied. This market segment will not seek (obtain) high quality advice. We can summarize these findings in the following proposition. Proposition 3 Competition between investment and universal banks leads to market seg- mentation. • Universal banks use tied deals to sell investment banking services for which its cost disadvantage is not too large to firms with high debt capacity (segment 1 and 2). This includes advice that is not profitable enough to be demanded from investment banks (segment 2). • Investment banks sell service that has a large impact on a client’s profits and for which their cost advantage is sufficiently large to firms with low debt capacity. 19
  20. 20. 5.1.2 Profitability of the fee business for universal banks In the absence of any constraints on debt levels in tied deals, entry of more than one com- mercial bank eliminates all profits (see Lemma (3)). If an upper limit of the debt level which can profitably provided exists, this extreme result vanishes: commercial banks have an incentive to enter the investment banking sector since they can, despite competition from other universal banks, expect positive profits. The mechanism behind this is the following. Suppose that more than one commercial bank has entered the market. With a high enough debt capacity of the firm under consid- eration, universal banks can guarantee incentive compatibility (i.e. ensuring that they will provide high quality of advice). This is the case if (6) holds for p = cU . Hence, competition among universal banks will lead to p = cU implying zero profits for this type of banks. In contrast, if the incentives stemming from debt are not sufficient to induce universal banks to choose high quality, universal banks earn a rent: p = cU makes it impossible to announce credibly the provision of high quality advice and p strictly exceeds cU . In a nutshell, a certain level of future profits has to be sustained to sustain incentives for high quality advice. That is, for projects/firms for which ˆ ˆ ∆(D − π) + (p∗ − cU )/r > cU > ∆(D − π) (8) and (5) hold, competition among several commercial banks and several investment banks implies that commercial banks outcompete pure investment banks. The equilibrium price is ˆ p = cU + r[cU − ∆(D − π)]. (9) This, in turn, leads to positive profits for commercial banks in the presence of competition among entering commercial banks. We can summarize: Proposition 4 Universal banks make positive profits on mandates with intermediate incen- tives stemming from the maximum debt in tied deals ((8) holds). For firms with high incen- 20
  21. 21. tives stemming from the maximum debt level, Bertrand competition results in zero profits for universal banks as tying alone is sufficient to sustain incentive. 5.2 Constraint on how many risky projects a commercial bank can finance Another reason why a universal bank cannot service the whole market is a constraint on how many risky debt it can provide. If the debt capacity constrains the number of tied deals that a commercial bank can offer, it might not be optimal to offer advice at a price below p∗ . There are several reason for capacity constraints for risky debt. Most notably, an universal bank’s costs of providing an increasing number of firms with risky debt are likely to increase due to diseconomies of scale in bank size, a bank’s cost of financial distress, and regulation. These costs impose an upper limit on a bank’s capacity for risky debt provision. In a wider sense, the existence of a capacity constraint can be interpreted as a representation if an upward- sloping marginal cost curve. The existence of a capacity constraint thereby delineates the extreme case: marginal costs become infinite at the capacity level. In order to explore this issue in more detail, suppose that two universal banks have entered the investment banking market and compete with investment banks. The universal banks’ costs of providing advice is cU > c. Both universal banks have limited capacity in ¯ the sense that universal bank i (i = 1, 2) is only able to provide Ni deals. In the third stage of our game firms will seek advice only if for a given p ∆(¯ − π) ≥ p π (10) holds. Furthermore, suppose that, e.g., due to differences in ∆ the impact of advice differs for different firms. This implies that the demand for (high-quality) advice is price-sensitive. Those firms/projects for which (10) holds for a given price will ask for advice whereas the remaining ones will not. We denote this price sensitive demand by M (p) with ∂M/∂p < 0 with the demand function denoted by p = M −1 . 21
  22. 22. Hence, we can state: ¯ ¯ Lemma 4 i) If M (p∗ ) ≥ N1 + N2 , then a unique pure strategy equilibrium exists in which universal banks charge p∗ or slightly lower. The universal banks will receive mandates from 2N firms, whereas the remaining ones are split among the investment banks at a price pIB = p∗ . ¯ ¯ ii) If M (p∗ ) < N1 + N2 and if a pure-strategy equilibrium exists, the two universal banks ¯ ¯ charge p1 = p2 = p ≡ M ax(p(N1 + N2 ), cU ). ˜ ¯ ¯ A pure-strategy equilibrium exists in this parameter range if Ni < Ri (Nj ) ≡ argmax(p(Ni + ¯ ˜ Nj ) − cU )Ni with j = 1, 2(i = j) or if Ni ≥ M ≡ M (cU ). In the former case only the universal banks receive mandates and make positive profits. ¯ For Ni ≥ M (cU ) Bertrand competition prevails implying zero profit levels for the universal banks. Proof : See Appendix. This Lemma implies that for the case of ”small” capacity levels we find coexistence of investment and universal banks. But even if universal banks are able to service the whole market, they might make positive profits. Only for very large capacities the conventional Bertrand result emerges. In our above discussion we have left out the regime of large capaci- ¯ ¯ ties, i.e. the one in which (M (cU ) > Ni > R(Nj )). In this range a mixed-strategy equilibrium emerges in the pricing game. Endogenizing the capacity game shows, however, that capacity levels corresponding to the Cournot-result, leading to a pure-strategy equilibrium, emerge (see Kreps and Scheinkmann (1982)). Therefore, it is for our purposes superfluous to consider this type of equilibrium.4 We can summarize: 4 This is quite intuitive. The Cournot capacities levels emerges from Max Ni (P (Ni +R(Nj ))−cU )−C(Ni ) whereby the last term denotes the costs of capacity installation. Hence, due to these costs (which are sunk ¯ ¯C in the price game) the Cournot capacity level (NiC = R1 (Nj )) are smaller than the capacity level necessary ¯ ¯ for a pure-strategy equilibrium (NiC < R(NiC )) a pure-strategy equilibrium results. 22
  23. 23. Proposition 5 For low enough capacities p = p∗ (or a price slightly below) emerges and universal and investment banks coexist. For intermediate capacity levels the universal banks divide the market among themselves, making positive profits. Only for very large capacities Bertrand competition prevails leading to zero profits. This reveals that even in the presence of competition among universal banks there is still room to benefit from the profitable investment banking business. It is quite obvious that these expected profits attract further market entry which might, however, be prevented by high costs of entry (i.e., cU ), build-up costs of capacity, limited resources (e.g., skilled ¯ bankers) etc. It also becomes obvious that in order to extract the maximum rent in the fee business, the universal bank wants to use its capacity from the provision of risky debt. Hence, it will ”not” provide firms with risky debt if theses firms do not also purchase fee business from the commercial bank’s investment banking branch. It will approach those firms first from which it expects the most promising fee-business (first market segment). 6 Empirical Implications This section summarizes the empirical implications. Implication 1: Tying is particularly important for universal banks that have a cost dis- advantage over specialized investment banks. Therefore, tying should be associated with the entry of commercial banks. At the time of entry cost disadvantages are arguably particularly pronounced. Tied deals allow commercial banks to enter the investment banking market and to compete against specialized investment banks in the profitable fee-business. Implication 2: Firms that obtain tied deals have higher debt levels. 23
  24. 24. A high debt capacity is required for debt to provide sufficient incentives. Therefore, tied deals are offered to firms with high debt capacity. Drucker and Puri (2003) provide supporting evidence. They find tying for firms with low credit rating. Ceteris paribus, a low credit rating results from a high debt level. Implication 3: Tying leads to more aggressive pricing of the commercial banks, thereby reducing the profitability of the fee business. Tying is associated with lower fees or lower interest (on the tied loan). In our framework lower interest rates and lower fees are perfectly interchangeable. Drucker and Puri (2003) find that tied deals are associated with lower fees or lower interest. Our model predicts that universal banks finance projects with ”high debt capacity”. These are projects/firms with low costs of high debt level (i.e., information insensitive assets, little risk shifting capabilities, low growth, e.g., hotel business) and projects with high debt financing for other reasons (e.g., project financing, LBOs). Implication 4: Tying can be used to provide advice when its marginal profitability is lower than the rent that induces specialized investment banks to provide high quality advice. Tying increases the incentive to return to the market more often and eventually undertake business with the same universal bank. The reason why firms which receive tied deals would return to the market more often compared to a situation with an absence of tied deals is that tied deals are cheaper, thereby making them more attractive. Marginal cost of using investment banking advice decrease with tying. But there is also an incentive to return to the same universal bank if a long-term loan has been provided which provides the necessary incentive for providing high-quality advice at lower price. Therefore, tying leads to repeated business (between the same universal bank and firm). Implication 5: Universal banks are more willing to provide risky debt (or credit lines) if this allows them to sell fee business. 24
  25. 25. Basically, we provide a rationale for the revival of the credit business which serves as a door opener for fee business. Therefore, our analysis leads us to expect that for commercial banks the entrance in the investment banking business should be ceteris paribus be associated with higher debt levels and more provision of riskier debt. Implication 6: Markets are segmented. Universal banks use tied deals to sell investment banking services for which its cost disadvantage is not too large to firms with high debt capacity. This includes advice that is not profitable enough to be demanded from investment banks. In contrast, investment banks sell service that has a large impact on a client’s profits and for which their cost advantage is sufficiently large to firms with low debt capacity. Robustness considerations and extensions Monetary incentives We completely abstracted from monetary incentives in investment banking services and assumed a flat fee. Of course, this assumption is extreme. Investment banking services are structured to provide the advisor with incentives to exert effort. For example, when a firm wants to issue bonds, it hires a bank as an underwriter who guarantees a fixed amount of proceeds that the issuing firm receives when the bonds are sold to investors. Thereby the underwriter assumes some of the market risk of placing the bond, providing the underwriter with incentives to provide effort to place the bonds. However, these incentives may not be sufficient: even when the total amount is guaranteed, there may be external effects that are not internalized, e.g., negative effects on the borrower’s reputation in the capital market for future transactions if the underwriter fails to place the borrower’s debt. In this case, the issuer’s reputation together with a high price paid for services provides additional incentives to exert effort. Hence, there is still a role for providing incentives through debt in tied deals even in this situation, where it is relatively easy to implement 25
  26. 26. explicit monetary incentives. For other types of services this will be more difficult. More generally, whenever incentive problems in investment banking services are not com- pletely internalized at zero rent for the investment banks, there can be a benefit to tying. The large profits earned in investment banking suggest that there is a rent to be earned. This rent might stem from imperfect competition but also from incentive problems. Chen and Ritter (2000) find a very high IPO spreads (7%) paid to investment banks in the United States. They argue that this spread results in a rent for underwriters and analyze potential justifications. Interestingly, they find that spreads on IPOs in other countries are much lower, including Japan and Europe, where universal banks dominate. Negotiating the terms of investment bank services and firms’ demand for tied deals In the present paper the market interaction was modelled as a take-it-or-leave-it- price-offer for investment banking services. In a different market model, where firms ne- gotiate with investment banks and universal banks, the ability of universal banks to offer tied deals can reduce the profitability of their investment banking business even when their constraint to provide risky debt is tight. Firms will explicitly seek tied deals to renegotiate the price for the advice. This is not possible without tying because the firm must always fear that if the investment bank accepts price cuts, it will do so for other customers as well to increase the amount of customers, jeopardizing incentives to provide high quality advice. This problem does not arise with tied deals where incentives are (also) provided through the debt repayment obligation. If the terms of investment banking services are negotiated between the bank and the client, firms’ want investment banks to also use credit because this allows to reduce the price of fee business. This is indeed a development that has be observed. As Cairns et al. (2001) argue: ”Still more worrying for investment banks is the fact that some clients now demand credit in return for M&A and underwriting business” (p 42). 26
  27. 27. Appendix Derivation of mixed strategy equilibrium – Lemma 3.3 With cU > p∗ an equilibrium in pure strategies does not exist. ¯ In the entry stage game, the two banks decide upon entering the market or not entering. The payoff of their simultaneous entry game are depicted as follows: Prob y 1-y 1/2 Enter Don’t Enter x Enter a,a a+b, 0 1-x Don’t Enter 0, a+b Vt+1 , Vt+1 whereby x and y denote the probabilities of the respective players to enter. If both enter the payoffs are p∗ − cU ≡ a < 0. If one bank enters and the other not, the entering firm receives ¯ p∗ − cU + (p∗ − cU ≡ a + b. If neither player enters the same game repeats next period itself ¯ with payoff Vt+1 . Since the two players are symmetric it suffices to look at player 1. Player one is indifferent between playing Enter and Don’t Enter if: y · a + (1 − y) · (a + b) = (1 − y) · Vt+1 (11) The expected payoff in the present period is: Vt = xya + x(1 − y)(a + b) + (1 − y)(1 − x)Vt+1 Since Vt = Vt+1 and due to symmetry (i.e. x = y) we obtain: x2 a + x(1 − x)(a + b) Vt = Vt+1 = (12) 1 − (1 − x)2 27
  28. 28. Plugging (12) into 11 yields after some calculations: a x=y =1+ b Proof of Lemma 4 i) Competition among investment banks will always lead to p∗ . Increasing the price to p∗ does not pay since it will lead to a discrete reduction in mandates while increasing the price only marginally. Increasing the price above p∗ even leads to a loss of all mandates. Undercutting below pU B does not alter the number of mandates and is therefore not profitable. Hence pU B = p∗ − and pI = p∗ is the unique pure-strategy equilibrium. ¯ ¯ ii) We consider the case of M (p∗ ) > N1 + N2 . For this case, we look at the equilibrium prices which result in a pure-strategy equilibrium (step 1). Then, in step 2 we derive necessary and sufficient conditions for a pure-strategy equilibrium emerge (step 2). Step 1: We start from the presumption that a pure strategy equilibrium emerges. First, suppose p1 = p2 > p. Hence, undercutting is profitable, since not all capacities are employed. ˜ Second, suppose p1 = p2 < p. If p = cU this is not an equilibrium since is involves losses ˜ ˜ ¯ ¯ (no production is hence better). If p = p(N1 + N2 ) rationing occurs. Hence, raising prices ˜ unilaterally therefore increases profits. Last, let’s consider pi < pj . Bank i has an incentive to increase its price as long as it is capacity constraint. i does not have an incentive to increase the price of it is the monopoly price (p∗ ). But then, j has an incentive to undercut. Hence, p1 = p2 = p emerges in the pure-strategy equilibrium. ˜ ¯ ¯ ¯ ¯ Step 2: Suppose Ni ≤ R(Nj ). In the pure-strategy equilibrium with p = p(N1 + N2 ) both ˜ sell their entire capacities. Hence, undercutting does not pay. Overbidding by i implies, by ¯ ¯ ¯ ¯ ¯ ¯ ¯ definition of R(Nj ) that p(Nj +R(Nj )) > p = p(N1 + N2 ), which, in turn, implies R(Nj ) < Ni ˜ ¯ ¯ thereby violating our initial presumption. Hence, Ni ≤ R(Nj ) is a sufficient condition for a ¯ ¯ ¯ ¯ ¯ ¯ pure-strategy equilibrium to emerge. With Ni > R(Nj ) we get p(Nj + R(Nj ) > p(Ni + Nj ) contradicting together with step 1 the existence of a pure-strategy equilibrium. 28
  29. 29. ¯ With Ni > M (cU ) we are back to the usual Bertrand case implying p1 = p2 = cU . The remaining statements follow straightforwardly. 29
  30. 30. References Allen, S., R.L. Clark, amd A. A. McDermed (1993), Pensions, Bonding, and Lifetime Jobs, Journal of Human Resources 28, 463-481. Benston, G. (1990), The Separation of Commercial and Investment Banking: The Glass- Steagall Act Revisited and Reconsidered, Oxford University Press. Cairns, A.J., J.A. Davidson, and M.L. Kisilevitz (2002), The Limits of Bank Convergence, The McKinsey Quarterly, July 2001, 41-51. Carlton, D.W., and M. Waldman (1998), The Strategic Use of Tying to Preserve and Create Market Power in Evolving Industries, RAND Journal of Economics 33(2), 194-220. Chemmanur, T.J., and P. Fulghieri (1994), Investment Bank Reputation, Information Pro- duction, and Financial Intermediation, Journal of Finance 49(1), 57-79. Chen, H.-C., and J. Ritter (2000), The Seven Percent Solution, Journal of Finance 55(3), 1105-1131. Choi, J. P. and C. Stefanidis (2001), Tying, investment, and the dynamic leverage theory, RAND Journal of Economics 32(1), 52-71. Drucker, S. and M. Puri (2003), Tying Knots: Lending to win underwriting business, Stanford University, mimeo. Kanatas, G., and J. Qi (1998), Underwriting by Commercial Banks: Incentive Conflicts, Scope Economies, and Project Quality, Journal of Money, Credit, and Banking 30(1), 119-133. Kanatas, G., J. Qi (2003), Integration of lending and underwriting: implications of scope economies, Journal of Finance 58, 1167-1191. 30
  31. 31. Kreps, D. and J. Scheinkman (1983), Quantity Precommitment and Bertrand Competition Yield Cournot Outcomes, Bell Journal of Economics 14, 326-337. Kroszner, R., and R. Rajan (1994), Is the Glas-Steagall Act justified? A study of the US experience with universal banking before 1933, American Economic Review 84, 810-832. Lazear, E. (1979), Why is There Mandatory Retirement? Journal of Political Economy 87(6), 1261-84. Puri, M. (1994), The long-term default performance of bank underwritten security issues, Journal of Banking and Finance 18, 397-418. Puri, M. (1999), Commercial banks as underwriters: implications for the going public process. Journal of Financial Economics 40, 373-401 Whinston, M.D. (1990), Tying, Foreclosure, and Exclusion, American Economic Review 80, 837-859. 31